Organisation internationale de normalisation

CDVS

The BoG on CDVS was chaired by Miroslaw Bober. The notes in this section were primarily recordings from the BoG, which met Monday-Thursday during announced times.

1. Review of input contributions

13.0.0.1.1.1.1.1.20m34644 A CDVS Pairwise Matching Experiment on a Dataset of Risso’s dolphins [Sabino METTA, Massimiliano ROSSO, Alberto MESSINA]

Presents the use of CDVA for the task recognition of "Rissos dolphins" with good results. Many mammalian species have distinctive natural marks allowing individuals to be identified from photos. This provides precious information for population size and population dynamic studies. Large photographic catalogues have been constructed for several parts of the world. Because the difficulty of matching animals' photographs increases with catalogue size, computer-assisted matching techniques are increasingly important.

The scope of this work is to use CDVS technology in a domain so far unexplored. This can be used for promotion.

13.0.0.1.1.1.1.1.21m34646 Request for minor updates to CDVS TM11 [Alberto MESSINA]

Proposal to add to the CVDS software an option which enables the output of the retrieval scores for the retrievied images. Accepted and already provided as a functionality in the proposal M35072.

13.0.0.1.1.1.1.1.22m35007 CDVS: TM11 Extraction and Retrieval Evaluation on ARM Architectures [Attilio Fiandrotti, Massimo Mattelliano, Alessandra Mosca, Giovanni Ballocca]

Benchmarks the CDVS extraction, matching and retrieval pipeline where client is using ARMv7 architecture (LG Nexus 5 mobile Phone). Dataset includes CDVS building dataset.

Extraction time is around 1s. Energy needed to extract is lower for the fast mode. Extraction of descriptors for 3500 images depletes the phone’s battery by approx. 20%. Matching pairs – 50,000 image pairs deplated 5-30% (depending on the descriptor size).